The influences of post-resignment services of the CEO’s on the R&D expenditures

(1)

MSc thesis

The influences of post-resignment services of the CEO’s on the R&D expenditures

Name: Omar Ebadi Student number: 11096659

Thesis supervisor: dhr. dr. G. (Giorgios) Georgakopoulos Date: June 24, 2017

Word count: 17032

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

Statement of Originality

This document is written by Omar Ebadi who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This study examined two major issues regarding managerial opportunistic behaviour during the last years at office. Firstly, this study expanded the prior literature by giving a better understanding of the

relationship between R&D expenditures and the horizon issue in the post-SOX period. Secondly, I examined whether post-resignment seats could serve as incentive/moderator in the horizon issue. By analysing the data of Forbes 500 companies between 2007 and 2015 I find that CEOs facing resignment show no sign of opportunistic managerial behaviour. Furthermore, the results suggest that

resignment seats indeed serve as a moderator in the horizon issue since the CEOs who provided post-resignment services have balanced R&D expenditures prior to departure compared to decreasing R&D expenditures of departed CEOs.

(4)

Various studies (Butler and Newman, 1989; Gibbons and Murphy, 1992; Murphy & Zimmerman, 1993; Edler, 2002; Heyden et al., 2015) examined the behaviour of the Chief executive officer (CEO). The most important reasons of those examinations are the bad publicity regarding earnings management of the CEO’s and the height of the CEO compensations. A report of Edmans (2016) shows that in 2013 CEO’s earned 350 times more than the average worker of the same company. While in 1980 the same ratio was 40 times more compared to an average worker.1_{Did CEO’s suddenly get more talent? Or does this}

problem arise due to agency issues? According to Edmans et al. (2009) CEO’s did not got more talent, instead talent became more important. Graham et al. (2013) finds that shareholders are willing to pay more in order to incentivize the CEO to operate in favour of the shareholders. According to Murphy (1986) CEO’s are worth every penny they get, as long as their talent or effort helps shareholders to increase their wealth.

As mentioned above, CEO’s came in bad publicity because of their earnings management tricks (Li & Zaiats, 2017). In order to fill their own pockets and to mislead shareholders, CEO’s use several methods. One of those methods is through manipulation of R&D. Normally companies invest in R&D in order to create future earnings. R&D is considered as an important tool to gain firm value over the years (Ettlie, 1998; Kim & Lyn, 1990). Some other researchers (Morbey, 1988; 1989) couldn’t find any connection between R&D intensity and future economic benefits. Kim and Lyn (1990) researched the same subject between US-owned companies and foreign owned companies who operated in the US. They found a relationship, but they indicated that there is also a difference in country of origin. Edler (2002) showed that R&D has become the cornerstone of corporate and business strategy. As it seems, R&D plays an important role in operating a business, making important decisions that have a big impact on firm value and eventually all of that can lead to higher profits.

Most of the time CEO’s are not involved in the R&D processes at all, while they are responsible for the R&D direction, rejecting and investing in R&D projects as well (Balkin, 2000). If we look at it from an accounting/process-oriented point of view, it’s pretty clear that there is a lack of segregation of duties. Shareholder give the CEO’s more decision space in order to give them the ability to invest in riskier projects. There is a difference in the willingness of taking risks between CEO’s and investors, because investors can put their own stock portfolio together and CEO’s are depended on one company. This makes it obvious that CEO’s are more risk averse than investors. To incentivize them through bonuses

(7)

and stock options, investors both encourage CEO’s to invest in riskier/high profitable projects as well as operate in line with investors interest (Croci, 2015).

In many cases CEO’s won’t fully leave the side of the shareholders after their retirement, they still remain on the board of directors (Quigley & Hambrick, 2012; Vancil, 1987). This phenomenon seems to be a trend in the recent years, especially in North America. Brickley et al. (1999) who provided evidence and collected large data of 277 retiring CEO’s between 1989 and 1993, focussed on whether the likelihood of post-retirement services of the current CEO is associated to the performances during the last years on job. According to Brickley et al. (1999) there is a positive relationship between post-retirement services of a CEO and performances while being on job. This means that there could be a link between the greater performance of the CEO during time served on the job and greater desirability of having the CEO on the board of directors after the resignation. On the other hand, Quigley & Hambrick (2012) finds that CEO’s take a position in the board after their resignation as symbolic honor. Lee (2007) who analysed two time windows during 1989-1993 and 1998-2002, showed that there is indeed a positive correlation between pre-retirement accounting performances and the number of outside directorship in the 1998-1993 sample, but this correlation vanishes in the more recent sample. He also concluded that stock price performances explains the movement of CEO in the board. Put it differently, in case the stock price performance of the CEO increases, the probability of post-retirement services of the CEO also increases. The latter seems to be present in both samples, but it got larger in the 1998-2002 sample. So far the evidence is related to performances as a whole. But which link can be made or is existing in the first place between the R&D expenditure during the final years of the CEO’s and their post-resignment position on the board?

1.2. Research question

The research question of this thesis is as following:

What is the influence of post-resignment services of the CEO’s remaining on board on the R&D expenditures in the final years at office?

To answer this question I have to analyse the differences in R&D expenditures during the final years at office between the situation where the CEO remains as chairman on board and the situation where the CEO will leave the company definitely. Like in some of the other researches (Cheng, 2004; Lin et al., 2012) I want to make some pre-assumptions. I assume that in case the CEO remains on board, the R&D expenditures will be slightly higher than in case the CEO will depart the company. A higher R&D

(8)

even if the CEO wants to leave the company and even if he can’t enjoy from the benefits related to investments, he will still invests in R&D which shows that he have prospects for the future. The research question including my assumption can be illustrated as follow.

Figure 1. Research question and assumption

R&D expenditure CEO leaves CEO remains on board Horizon Problem 2 1.3. Motivation of study

In the past, researchers examined the relationship between CEO compensations and earnings management (Tzioumis, 2008; Almadi et al., 2016), the effects of remaining CEO’s in board on performances while on job (Brickley et al., 1999; Evans et al., 2017) and opportunistic managerial behaviour regarding to R&D investments (Cheng, 2004; Tsao, 2014). Dechow & Sloan (1991) did a research about the horizon problem and the behaviour of the CEO during the last years at office. They have used the data of companies who had significant ongoing R&D activities during 1974-1988. Also Cheng (2004) did a research on R&D expenditures and their relationship with CEO compensation. Brickley et al. (1999) and Lee (2007) examined the effect of post-retirement services of a CEO on account and stock performances. There is a lot of research performed on the relationship of R&D investments and CEO compensations (Dechow & Sloan, 1991; Gibbons & Murphy, 1992; Murphy & Zimmerman, 1993; Wu & Tu, 2007; Heyden et al., 2015) and to a lesser degree on post-retirement services of CEO’s. Latter could be seen as limitations, due to the limited amount of information that is available about this subject To my knowledge there hasn’t been a research that examined the relationship between horizon problem, R&D expenditures and remaining CEO’s on board so far. Even though there hasn’t been a lot of research about this subject I want to take up a challenge and contribute to prior literature. First of all this means that with this study I will expand the prior literature by giving a better understanding of the relationship between R&D expenditures, CEO compensations and the horizon problem. Also Heyden et

2_{Executives who are likely to place little value on future earnings are those who have a short horizon} because they are expecting to leave their position in the near future. These executives have

incentives to boost their short-term earnings-related compensation by rejecting net-present-value

investments to develop intangible assets (Dechow & Sloan,1991). Researchers refer to this as the horizon problem.

(9)

al (2015) finds an association between horizon issues and the tendency of the CEO’s to manipulate R&D investments remain unexplored. Based on that, a further research is needed. Second, because of

incorporation of an additional variable (post-resignment services) it will generate new empirical evidences about whether CEO with shorter horizon have future prospects with the company or not. If the findings of Dchow & Sloan (1991) (CEO’s reduce their R&D expenditures during the last years at office) and findings of Brickley (1999) (likelihood of post-retirement services of the current CEO is associated to the performances during the last years on job) would be combined, it would mean that post-resignment services could be associated with greater R&D expenditures, compared to the departing scenario.

According to Wu & Tu (2007) who researched the relationship between CEO compensations and R&D expenditures, there is room for further research if additional variables such as CEO duality, director ownership or outside directors could be added in the research. They find that this kind of variables could influence the long-term investment (such as R&D). This is exactly what this study is examining. So far scientists examined those variables separately, but the effect of post-retirement services on R&D expenditures during the last years at offices is still unexplored. There is also another reason for further investigation of the horizon issues. So far the researches used the data before 2002. This means that the influence of the Sarbanes-Oxley (hereafter SOX) hasn’t been examined yet. According to Cianci et al (2011) SOX was not only intended to work as a preventive measure for financial fraud or ‘Shocks’ but also for poor governance practices and behaviour of management. Although the majority of this

legalisation is focused on accounting practices and board governance, there are also some parts related to executive compensation (for example Section 401 and 402). So by using post-SOX data, I can get

different results, compared to earlier studies.

At last but not least, if my assumption will be right, this study can give shareholders more information about the future prospects of the CEO during the last years on the job. As already mentioned, if the CEO will perform well during his job time, there could be a chance that the board members will offer him to join the board since he performed well and has a great knowledge of the company. The latter will automatically reduce agency problems since they got everyone on the same page. Therefore this research doesn’t only have theoretical contributions, but also practical contributions to agency issues.

1.4. Reading guide

The structure of this thesis is as follow: This study is divided in two parts, namely literature part (chapter 1 and 2) and empirical part (chapter 3-5). Chapter 2 consists of seven sections, whereby in the first two section the agency theory and optimal contracting theory is explained. Section 3 discusses the different compensation contract types and their relationship with R&D investments. In sections 4 and 5 is described what scientists find about the horizon of the CEO’s and their roll on the board after the

(10)

resignation. Chapter 2 ends with an overview table in which all used papers are summarized and the hypothesis development is described.

The empirical part starts at chapter 3 where the mythology, data sample and variables are discussed. Also a motivation for the sample selection and data cleaning is included in this chapter.

Chapter 4 consists of findings, results and main analyses. The empirical part ends with chapter 5 where the conclusion, limitation of the research and area for further research are formulated.

(11)

2. Literature review and hypothesis 2.1. Agency theory/problem

Agency theory could be marked as one of the most generally accepted theories when it comes to management and owner relationships. This theory describes the relationship between the agent (CEO) and the principal (shareholder/owners). Jensen & Meckling (1976) describe this relationship as follow: “A contract under which one or more persons (the principal(s)) engage another person (the agent) to perform some service on their behalf which involves delegating some decision making authority to the agent.”

In other words, principals hire an agent to act on behalf of the principal, (in this case) as a maximizer of the firm value. If we can assume that both parties are involved in this process as stock maximizers, we have a good reason to believe that in some cases the agent will have a different objective than the principal. According to Bonner & Sprinkle (2002) a CEO can have the intension to maximize his own profit above the shareholders profit. Frydman & Jenter (2010) finds that if managers are self-interested and shareholders do not know the best course of action, executives are likely to pursue their own well-being at the expense of the shareholder value. The conflicts of interests include the interest’s conflict between shareholders and debtholders and that between shareholders and management, both result in agency costs to the firm (Mu, 2016). Due to this conflict of interest, arises the so called principal-agent problem. In this research I only focus on conflicts of interest between managers and shareholders. Another important part of the descriptions of Jensen & Meckeling (1976) is delegation of decision rights. By giving the agent some decision rights, it becomes harder for the principal to fully monitor his actions. Beside the conflicts of interest, the principal-agent problem consists of a difference in willingness to take risks. As already mentioned, CEO’s are in general risk averse, because they are tied to one particular company. So even a small risk can bear big negative consequences. Shareholders on the other hand, can afford this risk, since they can manage their share portfolio by investing in different companies with different risk levels.

Decision processes as whole or events that will effect payoff can differ when two parties have different risk tolerances or risk appetites (Wilson, 1968). To align incentives, the literature has suggested two practical solutions, namely CEO compensation such as stock options and annual bonus plans and

monitoring of executives trough involvement of the Board of Directors (Michael & Pearce, 2004). During this research I will only focus on those two solutions, therefore other theoretical solutions will not be taken into consideration. Figure 2 shows how agency theory will be used in this research.

(12)

Agency theory

Principal- agent problem Different objective Solution Different risk appetite CEO compensation contracts Monitoring

Figure 2: Agency theory and solutions 2.2. Optimal contracting theory

As illustrated above, one of the solutions to the agency problem is the CEO compensation contracts. To analyse performance sensitivity of the CEO compensations, the classical framework of the optimal contracting can be used (Conyon & Murphy, 2000).Rooted in an economic view of incentives, this theory assumes that executives are in a strong position to effectuate better corporate performance, because of their discretionary powers and strategic oversight (Crossland & Hambrick, 2011). In order to improve performances, CEO’s have to put a lot of effort and take some risks, that in some of the cases can lead to negative results. To motivate the CEO’s or to provide them with satisfactory incentives, reward systems are invented that make CEO’s meet pre-specified performance targets (Baker et al., 1988). Reward system such as described in the optimal contracting theory will emerge from arm’s length contracting process between a CEO on one hand and the Board of directors on the other hand, in which competitive pressure from the markets and executive labor will push both parties to a performance dependent reward contract that is beneficial for all parties (Van Essen et al. 2012).

All over the world researchers (Melis et al., 2012; Van Essen, 2012) use optimal contracting theory when it comes to executives contracting. There are also some problems related to this theoretical model. First, the

(13)

majority of the researches relies on the data from the US, which means that the influence of the US culture forms the basis of the optimal contracting theory (Van Essen, 2012). Therefore this model doesn’t necessarily hold an international character. Second, optimal contracting theory focuses only on two particular parties, namely executives and board of directors. According to this theory, only these two parties participate in the compensation contracting process. However, there are also other parties such as creditors, blockholders and employees that can influence this process as well. During this research, lack of international character will be seen as a limitation. Other than that, this theory fits perfectly in the CEO and shareholder relationship and closes the gap in risk appetites and objectives of both parties.

2.3. CEO compensation contracts

To mitigate conflicts of interest between shareholders and managers, shareholder designed managerial compensation contracts. In general, the contract consists of two parts, namely, the fixed part and the variable part. The most common used contracts are: a basic salary, annual bonuses, stock options, restricted stocks and other long-term incentive arrangements (Lin et al. 2012). The variable part of the compensations can be divided into two different incentives groups, namely short-term incentives and long-term incentives (Dechow, 2006). To create a better picture of the forms, incentives type and short descriptions of it, the following table is created.

Component FP/VP Incentives

Basis Salary Fixed Short-term Fixed part that CEO gets no matter happens Annual bonuses Variable Short-term Variable part that CEO gets when targets

are met

Stock options Variable Long-term Valuable only when the price is above the exercise price. Can only be issued at the money.

Restricted stocks Variable Long-term Variable part that CEO gets when targets are met

Other long-term

incentive arrangements Variable Long-term Based on performances Table 1: Components of compensation contracts

The basic salary is the fixed part of the salary that an executive receives and is not tied to performances. This component of the salary is the only component that is fixed and the height of it will be determined in advance. The height of the CEO compensations are related to the investment opportunities, whereby firms with greater investment opportunities will have higher CEO compensation and vice versa (Gaver & Gaver, 1993). In other words, shareholders want to incentivize CEO’s of larger firms to a higher degree compared to smaller firms. The latter could be caused due to high likelihood of managerial opportunistic behaviour. On the other hand, the variable part of the salary is most of the time tied to performances

(14)

or/and targets. The literature (Cheng, 2004; Tsao, 2014) distinguishes also between the accounting-based compensation (ROA) and stock-based compensation (RET). There have always been a discussion about which compensation type is better and which has the most effects on alignment of incentives. In general both contracts have their advantages as well as disadvantages.

2.3.1. Accounting-based compensations

Firms can choose from several compensation contracts as illustrated in table 1. According to Almadi et al. (2016) there is high probability that self-serving will occur when a company largely relays on accounting-based performance measures in the CEO performances evaluation process. But why do companies still choose for accounting-based compensations? Brooks et al. (2001) finds that firms choose for this type of contracts, because these contacts are not separated from other contacts. By using this type, firms could use more than one compensation contracts in their remuneration package. Second, by adding this kind of contract as part of total compensation package, they could be more efficient by resolving agency issues. However, contracts of the CEO’s are for a finite time of life, while lifetime of a firm on the other hand is infinite. This means that CEO’s have much shorter horizon than the firm itself. Differences in horizons between the agent and the principal causes even more agency problems, since firms have a long-term horizon focus and CEO’s have a short-term focus. Cheng (2004) finds that CEO’s facing retirement will have more horizon issues since their horizon is shorter than the firms horizon and that’s why they become to a higher degree short-term oriented.

One of the primary functions of the accounting-based compensations is that those measures are reflections of the short-term decision making. This reflection function generates automatically a disadvantage by encouraging CEO to focus on short-term performances rather than long-term performances. Further, accounting-based compensations are backward looking measures, it makes management able to manipulate the numbers in order to influence the annual bonuses and maximize their private gains at the expense of shareholders (Almadi, 2016). According to Perez-Gonzalez (2006) accounting-based compensation are more sensitive to manipulations than stock-based compensations. The sensitivity level illustrates itself in the amount of accounting scandals in the past. Also Healy (1985) finds that accounting-based compensations makes managers able to manipulate earnings by transferring them between periods. For example, a manager can delay or even accelerate delivery of inventory at the end of the year. Doing so, he/she can shift the costs and as a result also the earnings. This negative incentive could be eliminated if the firm would provide the CEO with stock-based compensation (Wu & Tu, 2007).

2.3.2. Stock-based compensation

At the beginning of the 1980’s a lot of researches started to examine the effects of stock-based compensation on CEO performances (Haugen & Senbet, 1981; Eaton & Rosen 1982) In those years US

(15)

listed company firms started to add stock-based compensations in their total remuneration packages (Tzioumis, 2008). This type of contract is considered as one of the most effective variable compensation types. According to Jensen and Murphy (1990) this type of compensation contract incentivize CEOs the most, because CEO’s private wealth is directly linked to shareholders wealth. By giving them stock options CEOs are encouraged to focus less on annual bonuses and to a higher degree on long-term compensations. Doing so, there is a higher likelihood that CEOs will invest more in R&D in order to increase long-term profits and maximize firm value. Also Dechow & Sloan (1991) finds that providing executives with stock options will reduce opportunistic managerial behaviour prior to departure. Tzioumis, (2008) who researched the determinants of stock options as a component of executives compensations contracts of US listed firms between 1994 and 2004, came up with several reasons why firms like to use this type of contract:

1. Stock options reduce agency problem and align interest of CEOs with those of shareholders. By giving a CEO an option that is only exercisable when the share price exceeds exercise price, CEO will become less risk-averse and it will incentivize the CEO to maximize the share price (Guay, 1999).

2. Stock options will attract employees who are less risk-averse and are willing to put more effort in order to reach their goals. These kind of employees will rather stay on board in order to exercise their option than exiting their firm and losing the option (Oyer & Schaefer, 2005)

3. Firms with financial constraints will benefit from stock options since the cash flow wouldn’t be effected anymore (Core & Guay, 2001).

The findings of Tzioumis are not in line with prior literature (Dechow & Sloan, 1991; Gibbons and Murphy, 1992) since he showed that CEO age is negatively correlated with the adoption of stock options. The results showed in case the age of the CEO increases by 1 year, the probability of adoption of stock option will decrease by 0,4%. There are several possible explanations for this event. First, stock options have a long-term character, CEO’s who are facing a retirement are less likely to exercise their options because of their short horizon. Second, knowing that this option could serve as a retention mechanism, it sounds pretty clear that firms retention concern are greater for younger CEO’s who can get different offers from different firms, compared to CEO’s who are facing retirement. A third explanation according to Eaton & Rosen’s (1983) could be that age increase the level of risk-aversion, thus CEO’s could find stock options riskier compensation components compared to accounting-based compensations. A last explanation is related to the Gibbons & Murphy’s (1992) theory. They found that markets are assuming or expecting that CEO ability will decline when he or she approaches retirement. Because of the quality drop, they prefer to have to a higher degree accounting-based compensation than stock-based compensation.

(16)

2.3.3. R&D investments and CEO compensations

Compared to capital expenditures, R&D expenditures are riskier investments (Gosh et al., 2007) because the future economic benefits are at the date of the investment uncertain. But still for many innovative companies it’s a key factor to compete with their direct competitors (Delziel et al., 2011). Bhagat & Welch (1995) find that the central component of an inventive activity is the R&D investment. Edler (2002) showed in his research that R&D has become the cornerstone of corporate and business strategy.

If the R&D investment leads to future incomes and improves the competitive position of the company, why do CEO’s reject R&D investments in some cases? Dechow & Sloan (1991) researched CEO behaviour regarding R&D expenditures during the final years in office. Because of the common criticism that performances based on annual earnings encourage CEO’s to focus on short-term performances rather than long-term value creation, the main purpose of that research was to test whether CEO’s had the intension to improve their annual earnings by rejecting the R&D investment. Researches (Dechow & Sloan, 1991; Cheng, 2004; Gosh et al., 2007; Wu & Tu 2007; Heyden et al., 2015) referred to this issue as the “Horizon problem”. This kind of earnings management trough R&D investment is mainly caused by GAAP that require that all R&D investments have to be expensed at incurrence date. All the payoffs related to those R&D investments can only be recognized after the realization of it. Also Baber & Haggard (1987) find support for their hypothesis that managers manipulate R&D expenditures to “smooth” reported income.

Tsao et al. (2014), who used a sample of firms in Taiwan’s R&D intensive industries from 1996 to 2009, examined the relationship between CEO compensation and R&D investment in family owned companies. They concluded that firms can mitigate short-term investment behaviour by linking R&D expenditures to the executive compensation. Wu & Tu (2007) found new empirical evidence and supported the fact that there is a positive relationship between CEO compensations and R&D investments. In their research they focused on a sample of technology intensive firms who were listed in Standard & Poor’s 500 between 1995 and 2004. They showed that R&D spending is positively and significantly affected by CEO stock option pay. However, they also showed that CEO stock ownership has no significant effect on R&D investment. This means that there is difference in incentives alignment mechanism between stock ownership and option pay. CEO’s with stock ownership are less likely to invest in risky projects compared to CEO’s with stock options. Another interesting finding in this research is that the effect of CEO stock option pay on R&D investment is more prominent when company performances are particularly high and when a company has a large amount of absorbed and unabsorbed slack resources. There are also researchers who didn’t even find a relationship between the R&D investment and CEO compensations. Matsunaga (1995) scrutinized the accounting for employee stock options and the possibility of it to be used as a part of an income management strategy. He didn’t find any association

(17)

between R&D expenditures and the value of the employee stock options grands. Also Bizjak et al. (1993) concluded that R&D investments are negatively associated with CEO compensations.

The cross sectional correlation analysis of Johnson (1967) brings even more confusion in this mixed result situation, when he didn’t only detect no significant association between R&D and compensations, but also between R&D and future benefits at all. An assumption could be made regarding these mixed findings, whereby mixed findings could be attributed due to sample sizes, data quality and research era. Still an association between horizon issues and the tendency of the CEO’s to manipulate R&D investments remain unexplored (Heyden et al., 2015). These facts give a reason for further research between tendency to R&D manipulation and the horizon issues.

2.4. Horizon problem

In the past there have been some researches who scrutinized the association between R&D investment and career horizon issues (Dechow & Sloan, 1991; Cheng, 2004; Heyden et al., 2015). Cheng (2004) found in his research by presence of horizon problem a positive association between changes in R&D expenditures and changes in CEO total compensations. Similar results hold for changes related to CEO option compensations. Gibbons & Murphy (1992) found in their theoretical analysis that not only managerial actions are influenced by career horizon, but also incentives of current compensation contracts. Further they showed that when the career concerns are high, contractual incentives are low, and vice versa. A possible neutralizer to that problem could be an optimal compensation contract. Holmstrom (1982) on the other hand finds that reducing R&D investment trend in the horizon problem is caused due to the ability of managers. Managers who just started their career tend to work harder in their early years of work and less hard or not hard enough in later years.

As mentioned before Dechow & Sloan (1991) confirmed the assumption that the CEO does spend less on R&D during their last years in office. Even though the reduction of R&D expenditures is not associated with bad firm performance, these findings contradict the theoretical solution that was mentioned in paragraph 2.1. In other words, executive compensations don’t sufficiently encourage CEO’s to increase shareholders value or don’t fully eliminate conflict of interest between the agent and principal. Also Lundstrum (2002) finds that the R&D expenses are decreasing during the final years on job, but he concluded that these issues could be eliminated, in case the firm provides the executive during the final years with long-term compensation. The latter will reduce opportunistic managerial behaviour. Findings of Sougiannis (1994) declare this opportunistic managerial behaviour during the last years in office more clearly. He detected a lag of 3,4 years between the R&D expenditures and earnings period. Since the earnings period is 3,4 years, departing CEO’s do not benefit from these future payoffs associated with the R&D investments that they have to make. Besides that, because of the negative cash flows, their annual

(18)

bonuses that are tied to annual profits will decrease as well. This gives a CEO the motivation to reject the R&D investment in order to boost their annual bonuses.

On the other hand, Butler & Newman (1989) who also examined horizon issues regarding R&D investments, find no evidence about departing managers having incentives to manipulate R&D investments in order to increase short-term earnings. A possible reason for these results could be that the control mechanism that those firms use, works effectively even in the final years of the job, as a result they motivate the CEO not to take an opportunistic behaviour. Also Murphy & Zimmerman (1993) find no evidence that outgoing CEO’s exercise their discretion over accounting or investment variables to increase their earnings-based compensations. They find that declines in the R&D investment are better explained by the overall firm performances than by opportunistic managerial behaviour. Heyden et al. (2015) made an important addition to the literature by contextualizing the CEO disposition as a group level phenomenon and implying that the top management team plays a significant role in risk taking behaviour of the CEO in the final career stage. In other words, CEO’s interact in their final years to a higher degree with the top management team when they are shaping their R&D strategy. It’s not only the actions of the CEO’s that changes the R&D expenditures during the final years on job, top management teams have also influences. According to researchers one reason to inconsistent results or mixed findings between different studies could be that conventional approaches overlook an important fact that executives are not fully isolated agents when it comes to R&D decision making.

2.5. CEO remaining on board

So far associations between R&D investment and CEO compensations are pretty clear, but what is the effect of the remaining CEO in the board of directors on those variables? And why do CEO’s want to stay on board after their resignation? According to Lee (2007) one of the reasons why CEO’s provide their ex-firm with post-retirement services is, that the total size of outside directors is increased compared to recent years. Second, since ex-CEO’s have a lot of experience and enough skills, they can provide current CEO’s with clear and helpful advices. They have also a lot of knowledge to analyse and assess the strategy of the new CEO. Also Holmstrom (2004) gives some motives for the post-retirement services of the CEO’s. He finds that executives who remain on board are better placed to learn about firm’s strategy and understand better how current executives think about it. This information really matters when the company gets in trouble and the board members have to figure out whether the top management team including CEO have what the company need to get them out of the trouble. Even if it goes wrong, the board should be able to assess the actions of the CEO and the role of the external factors. The latter is in line with findings of Gibbons & Murphy (1992) who concluded that shareholder are not always certain about the ability of the new CEO. Shareholder find that CEO’s require some skills in order to pilot a

(19)

corporation. By having a CEO who knows everything about a firm, board can react timely and provide current management team with better advices.

Brickley et al. (1999) who provided evidences and collected large data about 277 retiring CEO’s between 1989 and 1993, focussed on whether the likelihood of post-retirement services of the current CEO is associated to the performances during the last years on job. In their research they analysed both continued services in their own board as well as services on other corporate boards. According to Birckley et al. (1999) there is a positive relationship between post-retirement services of a CEO and performances while on job. He even concluded that post-retirement services can have an effect on horizon issues. Brickley et al. (2000) showed that pre-retirement financial performances are significantly higher for the CEO who continue to serve on their former employees board after leaving the corner office. In their research they concluded that CEO’s who provide post-retirement services generate almost 11 percent higher abnormal stock returns and 2 percent higher accounting returns during their final years at office. They also looked at the best performed CEO’s (highest quartile of the financial performances) and found that those CEO’s are almost 30 percent more likely to provide post-retirement services compared to worst performing CEO’s (bottom quartile). Lee (2007) revisited the same subject and data as Brickley et al. (2000) but he added an additional time window of 1989-2002. He used more recent data to see what the changes are in the new era but also to compare the results to previous research. Lee (2007) found that the likelihood of ex-CEO who will serve as an chairmen after his resignation has an effect on the stock price performance (abnormal stock return). When the stock price performance increases by 25 percent, the probability that CEO will remain as a chairman or inside director during 1998-2002 goes up by 51 percent, compared to 11 percent of the prior research of Brickley et al. (2000). The results of this research are in line with findings of Holmstrom & Kaplan (2001) who pointed out that CEO pay to market performance sensitivity tenfold in the period between 1980 and 1999.

Butler & Newman (1989) examined a sample of firms from the Wall street journal between 1981 and 1982 for the announcements of the departures of the CEO’s. During the examination they focused on several variables as capital expenditures, R&D expenditures and finished goods inventory. The results were not consistent with assumption of CEO’s opportunistic behaviour. But on contrary, the results of retiring CEO’s could not be distinguished from non-retiring CEO’s. However they find that it’s difficult for the firm or the firm is even unable to penalize the managers after their departure, since his bad actions will appear the date after the departure. This means that when a CEO remains on board, there will be a possibility to penalize them for bad actions. The latter could also be related to the managerial behaviour regarding to R&D investment and horizon issues. Quigley & Hambrick (2012) on the other hand, finds that remaining position of the CEO does not have any influence at all and remaining CEO might take that position merely as a symbolic honor.

(20)

2.6. Overview used papers

As literature shows, researches have different opinions and mixed findings. Every researcher uses also different proxies and samples. To give an overview of all used papers with their used proxies and samples, the flowing table is created.

Table 2: Overview papers

R&D expenditures and CEO compensations

Author(s) Year Topic Sample size and

Country Proxies/controls variables

Dechow &

Sloan 1991 Relationship R&D expenditures and CEO compensations

58 firms; 1974-1988;

US R&D expense, Value of CEO stock options, Firm value, Changes in advertising and capital expenditures. Gibbons &

Murphy

1992 Effects of optimal incentive contracts when workers have career concerns

1292 CEO's

representing 785 firms and 6737 CEO-years; 1971-1989; US

Year’s sales, Stock price, CEO age, Tenure and total compensation. Murphy &

Zimmerman 1993 The effect of various financial measures surrounding CEO 1064 CEO in 599 Firms; 1977-1983, US R&D expenditures, Advertising costs, Accounting Accruals and Earnings, Sales, Assets, Stock prices.

Sougiannis 1994 The Accounting Based Valuation of Corporate R&D

573 Firms, 1975-1985, US

Effective tax rate, Net capital stock, Net value PPE, The inflation-adjusted book value of inventories; and The inflation-adjusted book value of recorded intangibles and other investments Bhagat &

Welch 1995 Explores the determinants of corporate R&D for U.S., Canadian, British,

6549 firm-years, 1985-1990, US, CA, UK, EUR.

Stock returns, Debt ratio, Operating cash flow and taxes.

Matsunaga 1995 The effects of financial reporting costs on the use of employee stock options

123 Firms, 1979-1989,

US Value of stock options, R&D expenditures and Firm size. Edler 2002 Changes in the strategic

management of technology of the world's most technology-intensive companies.

1991-1998 (Survey)Western-Europe, North-America and Japan

Annual sales volume, R&D intensity and Sales revenue from abroad.

Lundstrum 2002 Corporate investment myopia 10736 Firm-years;

1990-1993; US R&D expense, Market value of equity, Age of CEO, Long-term investment, Firm size

Cheng 2004 Effects of horizon and myopia problems on CEO

compensations

160 firms, 1984-1997, US

CEO Compensation, R&D Expenditures, Annual stock returns, ROE and RET.

Wu & Tu 2007 The relationship between CEO stock option pay and a firm's R&D spending

293 Firms, 1995-2004,

US R&D expenditure per employee, CEO stock option, CEO total current compensation and Firm performance. Gosh et al. 2007 Relationship between CEO

ownership and discretionary investments such as R&D and capital expenditures

9831 firm-years;

1994-1997; US R&D expenditure, Capital expenditure, Growth, Cash flow, Debt, and Size

Delziel et al. 2011 Influence of directors’ human and relational capital on firms’ R&D Spending

225 Firms; 2001-2003;

US R&D expense, Average age of directors, Debt to assets ratio, Inside and outside directors, Board size and Board tenure.

(21)

Tsao et al. 2014 Family ownership as a moderator between R&D investments and CEO compensation

2193 observations and 375 Firms; 1996-2009; Taiwan

CEO compensation, R&D investment, ROE, RET

Heyden et

al. 2015 Relationship between CEO career horizon, top management composition and R&D intensity.

100 Firms; 1998-2008;

US TMT Education, TMT Average Tenure, CEO Tenure, CEO Turnover, CEO Equity Ownership, CEO Base Salary and Firm Size.

CEO Compensations

Healy 1984 The effect of bonus schemes

on accounting decisions 239 company years (94 firms); 1964-1980 Discretionary and non-discretionary accruals, Bonus plan earnings, Account receivables and Inventory.

Brooks et al. 2001 The performance of firms before and after they adopt accounting-based performance plans

175 Firm; 1971-1980;

US OPRET and EPS

Dechow 2006 Asymmetric sensitivity of CEO cash compensation to stock returns

Based on other papers;

1993-2003 ROA, Annual Bonus and Stock options.

Perez-Gonzalez 2006 Impact of inherited control on firm performance 335 observations, 2002 Related and unrelated CEO succession, Market value, Abnormal returns, ROE and Net income to assets.

Tzioumis 2008 Determinants of stock option as a part of CEO

compensation

909 observations;

1992-2004; US CEO turnover, CEO age, Firm sales and CEO ownership Lin et al. 2012 Capital structure and executive

compensation contract design 1500 firms per year; 1992-2006; US Option grant, CEO characteristics, firm size, firm performance and ROA. Van Essen

et al. 2012 Relationship executive compensation and firm performance

332 studies, 29

countries Accounting performance and Total cash compensation. Almadi et al. 2016 CEO incentive compensation

and

earnings management

75 companies per country; 2005-2014; UK, Australia, Germany and Austria.

CEO incentive-based compensation, The level of investor

protection, Firm size, Firm growth, R&D and Firm financial leverage.

CEO Remaining

Butler &

Newman 1989 Effectiveness and decision making in an executive's final year with firm.

1730 Firms;

1981-1982; US Finished goods inventory, Capital investment expenditures and R&D expenditures.

Brickley et

al. 1999 The effects of post-retirement board services of a CEO. 277 CEO's representing 257 companies; 1991-1993; US

CEO's tenure, Tenure with firm, Total assets, Market value of equity, ROA, RTN and ARBET.

Lee 2007 What’s happened over the past 10 years to the

selection of retired CEOs as board members?

277 CEO's who retired and 250 Firms; 1989-1993 and 1998-2002; US

ROA, ARBET, Asset, CEO tenure and CEO age.

Quigley & Hambrick

2012 Effects of an CEO who stays on board on successor discretion, strategic change and performance.

181 successions; 1994-2006; US

ROA, TMT departures, TMT remaining, Tenure of predecessor and Age of CEO.

(22)

2.7. Hypothesis

The theories and literature that are used above, will form a basis for the hypotheses development and hypotheses testing. As mentioned above, there seems to be a relationship between R&D expenditures and managerial decisions during their final job years (Dechow & Sloan, 1991; Cheng, 2004). This relationship could be caused by manager’s incentives to build his own stake rather than increase firm value. For the first hypothesis I want to use the hypothesis of Cheng (2004). I would like to use their hypothesis, in order to analyse what the changes are in the post-SOX period but also to compare the results with his research. As mentioned before the most of the researchers used pre-SOX data regarding to this subject. By using the same hypothesis with different data, I will get new evidences on this topic. The first hypothesis is as follow:

H1: Ceteris paribus, changes in R&D expenditures and changes in CEO compensation are more strongly positively associated when the CEO is approaching resignment than when the CEO is not.

Another important component is the effect of the remaining CEO on the board after his resignation as chief executive. Since researchers found a relationship between post-resignation services and performances while on job, both R&D investment and opportunistic managerial behaviour could be affected if the CEO would remain on board after the resignation. At this stage of my study I want to make a pre-assumption. I expect that the post-resignation services will affect the performances during the job time positively. In other words, CEO’s who want to leave the company definitively, will reduce their R&D investment in a larger scale than remaining CEO’s. So post-resignation services will not only improve performances but also mitigate horizon issues. The hypothesis’s for this study is:

H2a: Ceteris paribus, CEO’s who stays on board after resignation will have less horizon issues than departing CEOs. H2b: Ceteris paribus, CEO’s who depart definitely will have more horizon issues than remaining CEOs

After testing this hypotheses I hope to contribute to prior literature and if my assumptions will be right than also provide shareholder more practical information.

(23)

3. Research methodology

This chapter explains what kind of methodology I used and how the data was gathered, cleaned and analysed. During this study I used the hypothetical deductive method. This method is a common used method when it comes to economics studies. This method consists of the following four steps:

1. Consider a new problem and establish research question with associated conjectures (Chapter 1) 2. Based on point 1, predictions needs to be deduced (Chapter 1-2)

3. Testing whether the sample is in line with prediction (Chapter 3-4) 4. Rejection or acceptance of the null hypothesis (Chapter 5)

3.1. Data and sample selection

To scrutinize this subject, the method of quantitative data research will be used. The research question will be answered by testing the hypotheses, which are formulated in paragraph 2.7. By gathering data from most commonly used databases for financial research purposes, regression analyses will be performed for all the variables. Wharton Research Data Service (hereafter WRDS) is the most appropriate source for the data gathering, related to this subject. The focus during this research will only be on the firms from the USA. Therefore COMPUSTAT (North-America) will be used. Unfortunately there isn’t a lot of public data about CEOs of other countries. Therefore to obtain data about CEOs, EXECUCOMP will be used. Furthermore, to prevent exchange rate issues during the analyses, only companies who are trading in US dollars will be added in the sample.

There are also requirements for useable data of CEOs. For instance, all CEOs who died or got fired will be excluded from the sample because those CEOs are not able to provide post-resignment services. It’s hard to find information around CEO dismissal, because this information most of the time affects share price negatively. So both shareholders and the company itself will try to hide that kind of information. Therefore, I gathered CEO sensitive information manually trough newspapers, annual reports and other reliable sources. CEOs who worked in a team or/and both resigned at the same financial year and joined the board at the same period will be excluded from the sample because it is hard to detect whose influence caused the results as they were at that particular year.

As other studies in the past (Dechow & Sloan 1991; Murphy & Zimmerman 1993; Cheng 2004), also this study will focus on Forbes 500 firms for several reasons. The conflicts between shareholder and management are more likely to incur by Forbes 500 firms compared to other firms (Dechow & Sloan, 1991). Forbes provides better and more complete data regarding CEO tenure and ages (Cheng, 2004). In order to select the most suitable firms, a list of Forbes 500 from CNN money will be used. Firms with the largest ongoing R&D activities will be added in the sample because of the higher probability of agency issues occurring, as mentioned above.

(24)

As mentioned before during this research the main focus lies on the post-SOX period. Before I started with data gathering I wanted to analyse data between 2002 and 2016 but because of the inability of WRDS, I had to shorten my time window from 2007 to 2015. Institutional Shareholder Service (ISS) allows us to use data between 2007 and 2015 so based on that my time window is shorted to these years. By merging datasets from ISS and Compustat (Fundamentels and Execucomp), I started with 6519 observations, 647 companies and 29.969 board members over the specified time window. After cleaning the sample I had 606 CEOs who provided post-resignment services. However, I discovered an error in the sample. In order to gather data regarding post-resignment services of the CEO, ISS is used. The standard procedure to find CEOs who provided post-resignment services is to look whether the “Date left company” is later than “Date left as CEO”. That would mean that the executive did resign as CEO of the firm, but still remained as board member of the same company. But an error seems to occur in the database regarding the leaving moment from the company. Most of the observations of “Date left company” were blank. Therefore, it wasn’t really clear whether the CEO really didn’t left company or the data was missing. This error occurred because I used recent data, so both options had the same chance of occurrence. In order to filter this error, I manually checked for every CEO the “Date left as CEO” and his function in ISS in the next 2 years. For example, if the “Date left as CEO” for CEO X was 2012 and “Date left company” was blank, but ISS showed that CEO X had no function in 2013 and 2014, that would mean that data was missing. Based on that, I changed each missed observation manually. Furthermore I deleted 3754 observations because those executives were current CEOs and didn’t leave their function. Based on the CEO criteria, I deleted 123 observations because tenure of CEO was less than 2 years and 631 observations were deleted because CEOs had left their function in 2015. The latter was deleted because it was not certain whether those CEOs would remain for 2 years (see 3.3.2). There is a possibility that these CEOs would leave the company in 2016 and based on that they are excluded from the sample. In addition, 395 observations were deleted because a CEO served in the same year as another CEO, 2 CEOs left at the same time or the company used the method of succession (Dechow & Sloan 1991). Since I analyze the performance of the CEO 2 years prior to departure, 112 observations were deleted because the CEO resigned during 2007-2008 and there wasn’t any data available on CEO performance before 2007.

Eventually after the final clean up round, my sample was reduced to 1407 observations over the time window, and 298 CEOs (99 CEOs who provided post-resignment services and 195 who departed a firm defiantly).

(25)

3.2. Variables

This paragraph explains which main variables (dependent and independent) and control variables will be used in this study.

3.2.1. Dependent variable

The dependent variable of this study is R&D expenditures (DV). There are different opinions and manners about measurement of this variable. In the past researchers calculated R&D expenditures through R&D per company sales (Dechow & Sloan, 1991; Daellenbach et al., 1999; Lundstrum 2002), R&D expenditures per company’s total assets (Naveen, 2006; Tsao, 2014), R&D expenditures per employee (Baysinger & Hoskisson, 1989; Wu & Tu, 2007) and as the ratio of R&D investments to firm revenues (Heyden et al. 2015). According to Danziel et al. (2011) it is better to avoid these R&D intensity measures because these kind of measures lead to uncertainty as to whether the effects are due to the numerator (R&D expenses) or the denominator (a proxy of firm size like number of employees or revenues). Based on that, this study will measure this variable through average R&D investment per year.

I will also look at the amount of the R&D expenditures during the final years on job before reassignment. Dechow & Sloan (1991) took 2 years before retirement, based on that, I will also look for the trend during the last 2 years at office. Doing so, I can draw a conclusion about probable decreasing investments. This variable will also be compared with the sample of CEOs who didn’t provide post-resignment services.

3.2.2. Independent variables

Post-resignment service (IDV1) is the independent variable that causes changes or variation on the R&D expenditures. A post-resignation place on board can serve as an incentive for the CEO to perform well. Also Brickley et al. (2002) find that post-resignation option is positively associated with current performances, and pre-retirement financial performances of CEOs who remain as board members are higher than departing CEOs. If we can assume that post-resignment service could serve as incentive, than a CEO should take that position for more than one year. Otherwise, his position would be merely as a symbolic honor (Quigley & Hambrick, 2012). Therefore, the post-resignation place would only be recognized if the CEO served at least two years in a row. I will create a dummy and will code this variable as 1 if the CEO will remain on board and provide post-resignation services and 0 if otherwise.

The horizon (IDV2) of the CEO affects the amount of the R&D expenditures during the last years on job. So it means that if a CEO has a shorter horizon he will spend less during the final years in order to boost his short-term compensations (Dechow & Sloan,1991; Lundstrum, 2002). According to Murphy (1999) CEOs typically retire around 64-66 years, this means that a higher age is associated with retirement as well as the likelihood of been asked to join the board. So based on that, this proxy equals 1 if the CEO reaches the age of 63 and otherwise 0.

(26)

Total compensation (IDV3) of a CEO could have influences on R&D expenditures prior to departure. If a CEO wants to boost his compensation prior to departure, then we should see an increasing trend during the last years at office. As the dependent variable, also this variable would be analysed 2 years prior to departure.

3.2.3. Control variables

During this study I will use the control variables that I find the most prominent and that fits the most into this research. The main focus lies on factors such as firm performance, board and CEO characteristics, risk taking level and pre and post-resignation services.

Accounting-based performance (CV1). According to Qui & Mak (2002) R&D expenditures are negatively associated with ROA. Performance of the CEO will be analysed during his last 2 years at office. To measure the accounting-based performance of the CEO, ROA will be used. This is in line with research of Brickley et al. (1999) and Lee (2007). However these researchers also used industry-adjusted ROA, and abnormal stock returns as measures of performance, but in my opinion this is unnecessary because those measures will be used in CV 2. As Tsao (2014), I will also use measurement as earnings before R&D spending and taxes divided by total assets at the beginning of the year.

Stock-based performances (CV2) will be measured like the study of Cheng (2004) and Tsao (2014) through the RET as stock returns over the year.

CEO tenure (CV3) will be measured by the number of years that a CEO worked in a firm. CEO tenure is in general associated with better managerial performances. According to Daellenbach et al. (1999) tenure enhances managers’ knowledge and makes them more capable of making investments in innovation that are necessary. This could imply that longer tenure makes CEOs use more trusted formulas with proven effectiveness and validity (Hambrick & Fukutomi, 1991; Miller & Shamsie, 2001). On the other hand, Hambrick & Mason (1984) find that CEOs, who have a longer tenure within a company, focus more on current products rather than new innovative terrain. Wiersema & Bantel (1992) suggest that CEOs with relatively shorter tenure take more risk. In other words, longer tenure could make a CEO more risk-averse. Also research of Baker & Mueller (2002) is in line with findings of Daellenbach et al. (1999) and finds that R&D spending increases with tenure of a CEO. The main reason for this is that CEOs may mold R&D investment to suit their own preferences. I this case, CEO’s larger tenure could increase R&D spending, CEOs performances while on job as well as probability of providing post-resignment services. Firm size (CV 4/FSIZE) is a common used variable in the past (Matsunaga, 1995; Heyden et al. 2015). According to Cohen & Klepper (1996) there is a positive relationship between R&D expenditures and the size of the firm. They even find that larger firms are in some ways advantaged in R&D investment processes compared to smaller firms. Hansen & Hill (1991) also found this relationship and measured this variable through natural log of firm revenues. However, firm size can also be measured based on total

(27)

number of employees (Heyden et al., 2015) and total assets (Matsunaga, 1995). During this study the total number of employees will be used to measure the firm size.

Percentage of outside directors (CV 5/%_OUTSIDE) controls how independent the board is. According to Dalziel et al. (2011) outside directors are those directors who are independent from the firm. The percentage of outside directors has a relationship with both the dependent as well as with the independent variable. As mentioned in paragraph 2.1 (figure 2) one of the main solutions to the agency problem is monitoring by the board of directors. Percentage of outside directors matters because if the board does not consist of enough outsiders, there is a high likelihood that the board wouldn’t be able to monitor the actions of a CEO unbiased and properly. Hence, I will measure this variable trough percentage of outside directors of total board size.

Gender (CV 6/GENDER) makes a difference when it comes to decision-making. Men are in general more focused on success than women and they attach relatively more value to success than woman. This could imply that male CEOs are willing to take more risk in order to achieve their goals. This is also in line with research of Byrnes et al. (1999) who showed that female participants are more risk-averse compared to the male participants. The research of Betz et al. (1989) gives a better explanation of gender by examining their values. He finds that the value of a particular gender has direct influences on decisions and interest. In case we can assume that male CEOs are more competitive than female CEOs, it sounds logical that male CEOs are more capable of breaking the rules (reject R&D investment) to achieve their goals (boost their short-term compensations) compared to female CEOs.

3.2.4. Summary table

To create a better picture of al variables and their measurement, the following table is made.

Variable Type Variable Variable code Measurement

R&D expenditures Dependent variable RD R&D investment per employee and average invested R&D

expenditures Post-resignment

service

Independent variable 1 POST Served at least two years in a row as board member

The horizon Independent variable 2 HORIZON CEO reaches age of 63 or higher Total compensation Independent variable 3 TOTAL_COM Total of compensation per year Accounting-based

performance Control variable 1 ACOUNT ROA (earnings before R&D spending and taxes divided by total assets at the beginning of the year) Stock-based

performances

Control variable 2 STOCK RET (stock returns over the year) CEO tenure Control variable 3 TENURE Number of years that CEO worked

in a firm

Firm size Control variable 4 FSIZE Based on total number of employees

(28)

Percentage of outside directors

Control variable 5 %_OUTSIDE Percentage of outside directors of total board size

Gender Control variable 6 GENDER Male or female

Table 3: Measurement of variables

3.3. Analysis

To analyse all the gathered data, there is need for STATA. After the clean-up round, the data was ready to use.

For this research the model of Cheng (2004) will be used. But this model doesn’t fit precisely into this study, so some of the variables are changed. For the examinations of the relationship between R&D expenditures and CEO’s horizon an ordinary least regression analyses will be performed with the following formula:

RDit = 𝛽o + 𝛽1 Horizonit + 𝛽2 TOTAL_COMPit + 𝛽3 ACCOUNTit (CV 1) + 𝛽4 STOCKit (CV2)+ 𝛽5

TENURE it (CV 3) + 𝛽6 FSIZEit (CV 4) + 𝛽7 OUTSIDEit (CV 5) + 𝛽8 GENDERit (CV 6) + 𝛽9

D_YEARit + 𝜀 it

i = index for firms; t = index for fiscal years;

In this model all the year dummies will be included. I will run this regression whereas R&D is the dependent variable and horizon the independent.

In order to test the effect of post-retirement services and reduction of R&D in the final years, also here the model of Cheng will be used. The regression seems as follow:

RDit = 𝛽o + 𝛽1 Horizonit + 𝛽2 POSTit + 𝛽3 TOTAL_COMPit + 𝛽4 ACCOUNTit (CV 1) + 𝛽5 STOCKit

(CV2)+ 𝛽5 TENURE it (CV 3) + 𝛽7 FSIZEit (CV 4) + 𝛽8 OUTSIDEit (CV 5) + 𝛽9 GENDERit (CV 6) +

𝛽10 Horizonit x POSTit +𝜀 it

This model is similar to model 1, but this model consists of an extra independent variable, which examines the influences of the resignation services and an interaction between horizon and post-resignment services. So this model is designed to test the H2.

(29)

4. Results 4.1. Pre-tests

In order to run a regression a few pre-tests are required to check whether the sample is suited for the regression. The normality test is an important check when it comes to regression analyses. The main reason for this is that if you are not satisfying the assumption for the OLS, there could be a possibility that non-normality would decrease the level of trustfulness of the results. Based on that, I tested whether the residuals are normally distributed. Figure 3 shows a bell shaped curve and confirms the normal distribution of the residuals with the highest values at the center of the histogram and decreasing values on the sides. In addition, Jarque-Bera is an excellent goodness of fit test when it comes to skewness and kurtosis. The Jarque-Bera test in figure 4 (See Appendix) shows a P-value of 0.85, which means that the null hypotheses of normality cannot be rejected.

After successfully conducting the normality test, I tested for multicollinearity. This phenomenon occurs when at least two or more independent variables correlate with each other strongly. In other words, collinearity is a linear association between variables in a model. Therefore I checked the variance inflation factor (VIF) value of all the variables in my model. In order to reject multicollinearity the VIF should be less than 5. However according to Chau & Gray (2002) a cutoff point of 10 is also an acceptable level. Figure 5 (see Appendix) shows that all VIF values are below 2, which means there is no sign of multicollinearity.

The last assumption for the regression is the absence of heteroscedasticity, which could invalidate the test of significance. Presence of heteroscedasticity could be confirmed if the standard deviation of variables is non-constant or when the variance of errors is constant. By rejecting the heteroscedasticity assumption, we automatically confirm presence of homoscedasticity. One way to detect heteroscedasticity is by making a plot of residuals. Figure 6 (see Appendix) shows that there isn’t any pattern observable and the dots are scattered randomly.

To be sure about homoscedasticity, I performed a Breush-Pagan test. The null hypothesis of this test is constant variance, which means that there isn’t any heteroscedasticity in the data. The p-value (0.81) of figure 7 (see Appendix) states that the null hypothesis cannot be rejected.

4.2. Descriptive statistics

In this research I identified CEO’s who provided post-resignment services for the same firm as where they worked as an executive. Figure 7 shows that the total sample consists of 294 CEO’s who left their position as CEO between 2009 and 2014. The major part of the CEO’s (66,33%) departed definitely and only 99 CEO’s (33,67%) stayed to provide post-retirement services after their resignation. The sample consists of only 3 female CEO’s and all of them were younger than 64 years, so they didn’t meet the horizon condition. As the table shows, the time window is reduced to 6 years and differs from the original

(30)

data as was mentioned in the methodology paragraph. Main reasons for this are the pre-resignment performances of the CEO’s prior to resignment and the post-resignment condition that a CEO should meet after the resignation date.

Table 3 Departures by year and age Panel A

CEO's who provided POST CEO's who departed Total Sample

< 64 64 and older (Horizon) Female < 64

2009 6 1 1 26 34 2010 11 4 1 31 47 2011 13 2 0 32 47 2012 22 0 0 32 54 2013 18 1 1 36 56 2014 17 1 0 38 56 Total 87 9 3 195 294

Panel B Departures by year and age

CEO's who provided POST CEO's who departed Total Sample < 64 64 and older (Horizon)

2009 7 1 26 34 2010 12 4 31 47 2011 13 2 32 47 2012 22 0 32 54 2013 19 1 36 56 2014 17 1 38 56 Total 90 9 195 294

I identified 294 CEOs who left their function during calendar years 2007-2015. This table reports which CEOs provided post-resignment services and met the horizon criteria.

I separated both genders as well as the group CEOs who met the horizon condition in order to analyze these groups separately. However, the sample of female CEO’s is too small (N=3) to base a conclusion on it. Therefore, I will only distinguish the CEOs who provided post-resignment services and were 64 and older and CEO’s who provided post-resignment services and were younger than 64. The latter is needed since I want to analyze whether the post-resignment position of a CEO will have significant influences on the pre-retirement performances and will help in controlling the horizon issues.

The influences of post-resignment services of the CEO’s on the R&D expenditures

Contents

Agency theory